Harnessing A.I. for Education

Introduction

Thank you for joining us for the program on harnessing AI for Education. I’m Marin Sahami, the chair of the CS department at Stanford University. It’s my distinct pleasure to introduce our main speaker today, Sal Khan. Sal is the founder and CEO of Khan Academy, a non-profit organization with the mission to provide free, world-class education to anyone, anywhere. They’ve partnered with over 500 school districts across the country, catering to more than 150 million registered users in over 190 countries, with lessons offered in more than 50 languages.

Sal is also the founder of the Khan Lab School in Mountain View, California, and co-founder of Schoolhouse World, offering free tutoring over Zoom. Furthering his impact, he co-founded Khan World School, a new online high school. Sal’s dedication to education began during his undergraduate years at MIT, where he developed math software for students with ADHD and tutored fourth and seventh graders in public schools. He holds three degrees from MIT and an MBA from Harvard Business School.

Background

Sal began Khan Academy in 2004, tutoring his cousin Nadia in math. What started as a family endeavor grew rapidly. His cousins began excelling, transitioning from remedial to advanced students. Soon, more family members and friends sought his tutoring, and Sal began creating online resources to scale this effort effectively. This included writing software for students to practice at their own pace and creating YouTube videos, which millions associate with Khan Academy today.

The Two Sigma Problem

In 1984, Benjamin Bloom, an education psychologist, identified the "two sigma problem," demonstrating the effectiveness of one-on-one tutoring in a mastery learning environment. Bloom’s study showed that personalized tutoring could move an average student well into the top decile, offering an improvement of two standard deviations. Khan Academy’s approach to personalized education has aimed to achieve this vision, striving to offer a similarly impactful educational experience.

Recent studies have validated Khan Academy’s impact, with evidence revealing significant improvements in student performance through accessible online resources.

Generative AI: A Game-Changer

Sal Khan emphasized the potential of generative AI in achieving even greater educational outcomes. Launching "Khanmigo," an AI tutor leveraging GPT-4, Khan Academy aims to provide an adaptive, personalized tutoring experience that goes beyond what traditional educators can offer.

Some features of Khanmigo include:

Interactive Maths Tutoring: Detects errors, provides tailored hints, and guides students through their reasoning.
Coding Assistance: Helps students understand their code, providing feedback and suggestions for improvement.
Literary Engagements: Allows students to chat with historical or literary characters, making learning interactive and appealing.
Debate Practice: Students can practice debates in a safe, judgment-free environment.
Essay Assistance: Offers feedback on structure, arguments, and style, ensuring students actively engage in the writing process.

Real-world Impact

In Newark, New Jersey, significant success has been seen since implementing Khanmigo. Students and teachers report high satisfaction and notable educational benefits. The AI tutor effectively mirrors a personal human tutor, aiding students in understanding, engaging, and excelling in their education.

Future Prospects

Ensuring the AI is culturally relevant and handles biases is a primary concern. Khan Academy actively tests and refines Khanmigo to address these issues.

While the cost of generative AI is currently high, Sal Khan is optimistic about cost reductions, making this technology more accessible. The vision is to level the playing field, ensuring all students, regardless of background, receive high-quality, personalized education.

Conclusion

The future of education lies in harnessing AI to provide personalized, engaging, and effective learning experiences. Khan Academy, through innovations like Khanmigo, aims to democratize education, meeting students where they are and helping them reach new heights. Sal Khan's approach reflects an unwavering commitment to intimacy, personalization, and equity in education.

Keywords

Education
AI
Generative AI
Khan Academy
Tutoring
Personalized Learning
Mastery Learning

FAQ

What is the "two sigma problem"? The "two sigma problem" identified by Benjamin Bloom in 1984 shows that one-on-one tutoring in a mastery learning environment can significantly improve student performance, offering an improvement of two standard deviations.

What is Khanmigo? Khanmigo is an AI tutor powered by GPT-4 introduced by Khan Academy. It provides personalized assistance in various subjects, including math, coding, literature, and more, aiming to mirror the effectiveness of one-on-one human tutoring.

How does Khanmigo ensure it doesn’t contribute to cheating? Khanmigo is designed with safeguards like conversation monitoring and AI supervision to ensure it guides students through learning without providing direct answers.

Can Khanmigo be used in non-academic areas? Yes, Khanmigo, while focused on academic subjects, also offers support in areas like career coaching and dealing with issues like procrastination, utilizing principles like cognitive behavioral therapy.

Is there a plan to make generative AI more accessible? Yes, while the current computational costs are high, Khan Academy is working towards reducing these costs to make the technology more accessible to a broader range of students.

How does Khan Academy address biases in AI? Khan Academy actively tests and refines Khanmigo to handle biases, employing red hat testing and ensuring the AI learns and adapts from diverse community inputs.